Slack-Bot-LLM-Serverless

A Slack bot with LLM support, using the Slack Events API, AWS Lambda, written in Python, utilizing the Serverless framework.

Initial Development Setup

Before you start, make sure to install all requirements!

Install serverless

brew install serverless

serverless plugin install -n serverless-pydeps

Creating virtual environment + install dependencies

python -m venv venv
source venv/bin/activate  # For Unix-based systems
venv\Scripts\activate  # For Windows

pip install -r requirements.txt
pip install -r requirements-dev.txt # For local testing

Do not forget to also add the venv in your pycharm interpreter: Settings -> Python Interpreter -> Add Interpreter -> Virtualenv Environment -> Existing

Management of Python Dependencies in Lambda

This project uses the serverless-pydeps plugin to manage Python dependencies. You can add dependencies to the requirements.txt file located in the project root directory. During deployment, the plugin will package the dependencies with the function.

IAM permission setup

The serverless.yaml file contains IAM role statements required for the Lambda function. The IAM role allows the function to access AWS services such as CloudWatch Logs, and AWS Systems Manager Parameter Store. The relevant role with all defined permissions is created during deployment.

Deploying the Function

sls deploy --verbose --config serverless.yaml --stage Dev

Stage needs the environment, you want to deploy to. For example "Dev", "Test" and "Prod"

This will package and deploy your Lambda function to AWS.

Running the lambda function locally

This can be useful for debugging! Disclaimer: This will use your local aws cli user, make sure you have sufficient permissions!

sls invoke local -f main --stage Dev

Replace main with the name of your function if you have changed it in the serverless.yaml file.

Monitoring and Logging

The Lambda function is configured to log messages to Amazon CloudWatch Logs. You can view the logs in the AWS Management Console or use the AWS CLI to retrieve log events.

One amazing feature of this setup is to view the live logs of deployed functions:

sls logs -f main -t --stage Dev 

Replace "main" with the name of your function if you have changed it in the serverless.yaml file. The -t flag will enable tailing the logs in real-time.

The utils/logger.py module contains the log_to_aws function, which you can use to log messages of various log levels ( e.g., LogLevel.INFO, LogLevel.DEBUG, etc.). You can import and use this function in your Lambda handler to log custom messages.

Removing the Deployed Function

1. Open the AWS web console
2. Browse to CloudFormation
3. Find your "Stack" and delete it

Troubleshooting Common Issues

If you encounter issues during deployment or while running your Lambda function, consider the following common causes and solutions:

1. Missing or incorrect IAM permissions: Ensure that the IAM role statements in the serverless.yaml file include all necessary permissions for the AWS services your function is using.
2. Dependency issues: Make sure that all required dependencies are listed in the requirements.txt file and installed in your virtual environment. Verify that the serverless-pydeps plugin is properly configured in the serverless.yaml file.
3. Invalid event sources: Double-check the event sources specified in the serverless.yaml file for your Lambda function. Make sure they are correctly configured and use the appropriate syntax for the Serverless Framework.
4. Errors in your Lambda function: Check the logs in Amazon CloudWatch Logs for any error messages or stack traces that can help you debug issues in your Lambda function. Use the log_to_aws function from the utils/logger.py module to log custom messages that can help you identify issues.
5. Resource limits: Monitor the usage of resources such as memory, CPU, and execution time for your Lambda function. Adjust the settings in the serverless.yaml file as needed to ensure your function has enough resources to run successfully.

Testing

Our default for testing is the python package pytest.

Run existing tests

To run all tests, use the following command in your project root:

pytest

To run specific tests based on their markers, use the following command:

pytest -m <marker_name>

# Current tests have the tags "unit" and "integration" (They are defined in our pytest.ini)

# To run only unit tests, just run:
pytest -m unit

Add new tests

1. Create a test file for your respective handler in our "tests" folder. The naming convention is test_ ( main.py -> test_main.py)
2. In this file you create a function and add the Python decorator (annotations) @pytest.mark.<marker_name>. We currently have two markers.

Todos's

• Formatting breaks some links, do the same thing as with code blocks
• Improve websearch by generating a more precise question for the search(include history data to come up with a search query)
• Add marker for web search responses "[Web-Search]"
• Intention implementation: Confluence content search(https://developer.atlassian.com/cloud/confluence/rest/v1/api-group-search/)
• Refactor / cleanup "async_processor.py"
• Intention implementation: Image generation
• Intention implementation: create ticket
• Implement special commands for prompts(move web search into this)?
• Add support for external knowledgebase(e.g. Confluence)
• full rewrite Readme.md
• Introduce env var file vs Parameterstore reading + fix Env var check in main
• Switch to Slack SDK?
• Add support for other LLMs.

Support me ❤️ ⭐ 💸

If this project provided value, and you want to give something back, you can give the repo a star or support by buying me a coffee.